Uninstalling rancher2 objects from EKS etcd cluster

I’ve installed rancher2 into a testing cluster and am iterating on some tooling for our environment. I installed it using the addon manifest from 3-node-externalssl-recognizedca.yml so it lives in the managed cluster.

When deleting it, via kubectl delete -f manifest or deleting the cattle-system namespace, data seems to be left in etcd or somewhere.

After a re-apply of the manifest, it comes back knowing the administrator password and other information that it shouldn’t know.

Can I evict all rancher data out of the cluster so I can start fresh? It is EKS so I don’t have visibility into etcd, which makes me wonder how rancher does.

Ah! Looks like it is all stored in CRDs and some annotations/labels.

I got some namespaces wedged in Terminating because of the finalizers, I patched them to null the finalizer then removed all the CRDs with cattle.io

kubectl get crd --output=custom-columns=NAME:.metadata.name --no-headers | grep cattle.io | xargs -I% kubectl delete crd %

Had some user-XXX and project-xxx namespaces and local to clean up.

I don’t know if this is what we are supposed to do, but when I re-applied rancher, it started at admin password and asked to set the URL again.